Simple Minimax Optimal Byzantine Robust Algorithm for Nonconvex Objectives with Uniform Gradient Heterogeneity.
Tomoya MurataKenta NiwaTakumi FukamiIifan TyouPublished in: ICLR (2024)
Keyphrases
- worst case
- dynamic programming
- detection algorithm
- computationally efficient
- cost function
- exhaustive search
- optimal solution
- k means
- closed form
- np hard
- objective function
- optimization problems
- gradient information
- linear programming
- expectation maximization
- search space
- neural network
- matching algorithm
- preprocessing
- convex optimization
- parameter tuning
- game tree
- gradient vectors
- optimal path
- optimal parameters
- optimal strategy
- particle swarm optimization
- computational cost
- lower bound